# How Quantum Computers Learn to Sort Data > This patent describes a method for training a data classifier using a quantum computer, where classical AI helps prepare the data and interpret the quantum results to improve the quantum system's ability to sort information. - **Patent:** US 20240296359 - **Original title:** Classification using quantum neural networks - **Owner:** Google - **Status:** Active - **Times cited:** 0 - **Field:** quantum_computing, ai_ml, software, telecommunications, semiconductors, biotech ## What it does The patent details a method for training a classifier that runs on a quantum computer. First, a classical artificial neural network prepares a set of quantum bits (qubits) in an input state with known categories (Claim 1). These qubits then go through a series of "parameterized quantum gates" — like adjustable switches — which change their quantum state. Another classical artificial neural network then reads the final state of specific "readout qubits" to guess the category of the initial data. This guess is compared to the known category, and the quantum gates' adjustable settings (parameters) are tweaked to make future guesses more accurate (Claim 1, 5). This training process repeats until the classifier performs well (Claim 2). Once trained, the quantum computer can then classify new, unknown data by applying the learned quantum gates and interpreting the readout with the second classical network (Claim 15). For example, a quantum computer could be trained to distinguish between quantum states representing different types of particles, with classical AI assisting in the setup and interpretation. ## What it does NOT cover - Training a quantum classifier without using any classical artificial neural networks for preparing the input state or determining the predicted classification (Claim 1). - Classification methods that do not involve parameterized quantum gates (Claim 1). - Quantum machine learning tasks other than classification, such as regression or generative models. - Training methods that update parameters of a classical neural network based on quantum computations, rather than updating parameters of quantum gates. - Classification systems where the input state is prepared purely classically without being encoded into qubits. ## The clever bit The novelty lies in the hybrid approach, specifically using two distinct classical artificial neural networks to assist the quantum computer during the training process: one for preparing the input quantum state and another for interpreting the quantum readout to determine the predicted classification. This bridges the gap between classical data and quantum computation, making the quantum classifier trainable and usable. ## Real-world examples 1. Google's Sycamore processor running classification algorithms 2. IBM Quantum Experience exploring hybrid quantum-classical machine learning 3. Quantum-enhanced image recognition 4. Quantum algorithms for financial fraud detection 5. Drug discovery applications for classifying molecular structures ## Why it matters Quantum computers hold promise for solving problems too complex for traditional computers, including advanced machine learning. This patent addresses a fundamental challenge in quantum machine learning: how to effectively train these new types of classifiers. By combining classical AI with quantum processing, it aims to make quantum computers practical for tasks like pattern recognition and data sorting, potentially speeding up drug discovery, materials science, and financial modeling. ## Frequently asked questions ### What does How Quantum Computers Learn to Sort Data cover? This patent describes a method for training a data classifier using a quantum computer, where classical AI helps prepare the data and interpret the quantum results to improve the quantum system's ability to sort information. ### Who owns patent US 20240296359? This patent is owned by Google. ### When does this patent expire? This patent is expected to expire on April 26, 2044, when the invention enters the public domain. ### What problem does this patent solve? Quantum computers hold promise for solving problems too complex for traditional computers, including advanced machine learning. This patent addresses a fundamental challenge in quantum machine learning: how to effectively train these new types of classifiers. By combining classical AI with quantum processing, it aims to make quantum computers practical for tasks like pattern recognition and data sorting, potentially speeding up drug discovery, materials science, and financial modeling. ### What does this patent NOT cover? Training a quantum classifier without using any classical artificial neural networks for preparing the input state or determining the predicted classification (Claim 1). **Full plain-English explainer:** https://patentbrief.org/patent/us/20240296359/classification-using-quantum-neural-networks **Original patent:** https://patents.google.com/patent/US20240296359 --- _Source: PatentBrief — https://patentbrief.org. Patent facts are from public records; the plain-English explanation is PatentBrief's._ ## Related patents Semantically similar inventions in the PatentBrief corpus: - [How Quantum Computers Use Special Devices to Control Qubits](https://patentbrief.org/patent/us/9256834/quantum-computers-having-partial-interferometric-quantum-gates) — This patent describes a way for quantum computers to control special quantum bits, called topological qubits, using a "partial interferometric device" to perform specific operations. - [How D-Wave Clears Magnetic Noise in Quantum Computers](https://patentbrief.org/patent/us/11295225/superconducting-quantum-processor-and-method-of-operating-same) — A method for improving quantum computer accuracy by actively clearing out magnetic interference that builds up during calculations. - [How AI Systems Learn to Predict and Act Simultaneously](https://patentbrief.org/patent/us/11170293/alphago-policy-and-value-networks) — A method for training AI models that combines supervised learning for prediction with reinforcement learning for decision-making in a single, coordinated system. - [How Projection Neural Networks Speed Up AI Predictions](https://patentbrief.org/patent/us/11544573/llama-large-language-model-architecture) — A method for making artificial intelligence models faster and more efficient by using fixed, non-trainable projections to simplify complex data before processing. - [Making AI Smarter by Focusing on Unsure 'Nodes'](https://patentbrief.org/patent/us/12423586/training-nodes-of-a-neural-network-to-be-decisive) — This 2025 patent from D5AI LLC describes a way to train AI models more effectively by boosting the learning signal for 'nodes' that aren't making clear decisions on data.